[Homer-users] Trouble with Homer2: Using Artifact marking and GLM (Drift) together, sets CCs to zero???

[Juliette Selb]

Hi Marina,

The GLM analysis only includes the time points without motion artifacts (tIncAuto).

When you display Concentrations, what is shown is not directly the real concentration time traces, but the reconstructed concentrations calculated from the GLM model (i.e. HRF times design matrix + residual). Because the model is not applied on segments with motion artifacts, these appear as 0. Are you using motion corrections? It could help, or maybe what you see is from remaining motion artifacts after correction.

We could modify the code to display the "real" concentrations instead of the "reconstructed" concentrations, but you wouldn't be able to visualize the short separation regression for instance.

Hope this makes sense,
Juliette




-----Original Message-----
From: homer-users-bounces at nmr.mgh.harvard.edu [mailto:homer-users-bounces at nmr.mgh.harvard.edu] On Behalf Of Marina Winkler
Sent: Monday, May 23, 2016 4:02 AM
To: homer-users <homer-users at nmr.mgh.harvard.edu>
Cc: Maren Grigutsch <grigu at cbs.mpg.de>; Katherine Perdue <katherine.perdue at childrens.harvard.edu>; maennel <maennel at cbs.mpg.de>
Subject: [Homer-users] Trouble with Homer2: Using Artifact marking and GLM (Drift) together, sets CCs to zero???

Dear Mailinglist,

I am having (again) some serious trouble with the Homer2 Toolbox. I have the newest version of Homer2 (2.1), but this trouble I also experienced with an older version of Homer. 

Whenever I use the artifact detection and the GLM (drift) together, my concentration changes (CC's) get totally messed up -> areas containing artifact markings get set to zeros in the CC's.  

Homer2 does this not only for my infant data but also for the sample data (which did not contain artifacts, but illustration purposes I set the sensitivity very high, to have some areas marked)... -> Pictures attached.

What I tried so far:

-> Using artifact marking together with a PCA or PCA recursive and GLM (drift) - still had zero's in the CC's.
-> Using another GLM function (GLM Drift SS) - problem persisted.
-> Just artifact detection and no GLM - no zero's in CC's.
-> not using artifact marking, but just Filter and GLM - no zero's in CC's.  

Do your Homer2 toolboxes do this as well, or is this only mine? 

Might this be a problem in the parameters/functions I use? If so, which alternative functions should I use instead? 
I need a BPF, an automatic artifact detection and correction by PCA and a GLM.   

Where in the scripts or structures/variables might this problem be? 
I have a fresh Homer2 install, without any self-written code in it.   

Thank you very much,
Marina


__________________________________________________________________________
 Marina Winkler
 PhD student IMPRS NeuroCom
 Department of Neuropsychology
 Max Planck Institute for Human Cognitive and Brain Sciences
 
 Office: +49 341 9940-2586
 Mobile: +49 1578-388 1089
 E-mail: winklerm at cbs.mpg.de 
 
 MPI-CBS Leipzig
 Stephanstraße 1a     
 04103 Leipzig, Germany  

               
__________________________________________________________________________